Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Conduction II: Spreading Resistance Addendum 1

CHAPTER 12 ADDENDUM

Conduction II: Spreading Resistance Addendum


This addendum supplements Chapter 12 by including Chapter 12 topics that were in the first
edition, but not in this second edition. The following topics are no doubt of a sufficiently complex
mathematical nature, that many readers will choose not to study them, but for those who might be
interested, they are presented here.

12.10 APPLICATION EXAMPLE: IC CHIP ON AN ALUMINA SUBSTRATE

A single integrated circuit chip on an alumina-ceramic substrate with an adiabatic source plane
is illustrated in Figure 12.34. The dimensions are a = 1.0 in., b = 1.0 in., t = 0.04 in., ∆x =0.05 in.
and ∆y =0.05 in. with a thermal conductivity of k=1.0 W/(in. oC). We shall look at two different
situations where the first is for a convective heat transfer coefficient h = 0.005 W/(in.2 oC). The first
step is to calculate the required dimensionless variables.
∆x a = ∆y a = .05 in. 1.0 in. = 0.05,τ = t a = 0.04 in. 1.0 in. = 0.04
For this first case, the Biot ⋅ τ number is Biot ⋅τ = ht k = ( 0.005 )( 0.04 ) 1.0 = 2 × 10−4
The problem solution follows. Using Figures 12.14 and 12.23, we find that ψ Sp = 0.78 and
ψ Sp = 0.69 . The maximum total resistance is then found to be
R = RU + RSp = (1 hab ) + ( t kab ) + ψ Sp k ∆x∆y
1 0.0.04 0.78
= + +
( 0.005)(1.0 ) (1.0 ) (1.0 ) (1.0 ) ( 0.05)( 0.05)
2 2

R = 200.00 + 0.04 + 15.60 = 215.64 o C W


and the average value is
R = RU + RAve − Sp = (1 hab ) + ( t kab ) + ψ Sp k ∆x∆y
1 0.0.04 0.69
= + +
( 0.005)(1.0 ) (1.0 )(1.0 ) (1.0 ) ( 0.05)( 0.05)
2 2

R = 200.00 + 0.04 + 13.80 = 213.84 o C W


Both the maximum and average total resistances are dominated by the convection term.

Figure 12.34. Application Example 12.10: Geometry for IC chip on ceramic.


1
2 Thermal Computations for Electronics

The second case that we consider is when the convecting surface becomes that of an infinite
heat sink so that h → ∞ and Biot ⋅τ → ∞ . Once again using Figures 12.14 and 12.23, we find that
=ψ Sp 0.43and
= ψ Sp 0.34 ,
0.0.04 0.43
R = RU + RSp = (1 hab ) + ( t kab ) + ψ Sp k ∆x∆y = 0.0 + +
(1.0 )(1.0 ) (1.0 ) ( 0.05)( 0.05)
2

R = 0.0 + 0.04 + 8.60 = 8.64 o C W


and
0.0.04 0.34
R = RU + RAve − Sp = (1 hab ) + ( t kab ) + ψ Sp k ∆x∆y = 0.0 + +
(1.0 )(1.0 ) (1.0 ) ( 0.05)( 0.05)
2

o
R = 0.0 + 0.04 + 6.80 = 6.84 C W
where we see in this case that the spreading resistance now dominates the problem. The uniform
conduction contribution is trivial in both the finite and infinite h situations.
Unit-source thermal contours are plotted in Figure 12.35 for finite and infinite h, where you
should note that the left and right edges of the substrate are truncated. The finite h shows the strong
tendency of the heat flux vectors to spread. And as expected, the infinite h situation shows that the
heat flux vectors tend to be largely directed downward toward the heat sinking surface.
Spreading theory.
Left and right edges cropped.
Source extent
← →

0.4 0.5 0.6


h = 0.005 W/(in.2 C)
o

Figure 12.35 (a). Application Example 12.9: Unit-source temperature contours taken through center plane.

Spreading theory.
Left and right edges cropped.
Source extent→

0.4 0.5 0.6


h →∞
Figure 12.35 (b). Application Example 12.9: Unit-source temperature contours taken through center plane.

12.11 RECTANGULAR SOURCE ON RECTANGULAR, FINITE-MEDIA


WITH TWO CONVECTING SURFACES: THEORY

Now we consider an extension to the work in Section 12.6. The new part of the problem is
that we add Newtonian cooling to the surface at z = 0 via a heat transfer coefficient h1 in an ambi-
ent temperature TA1 and the heat transfer coefficient at the opposing surface is renamed as h2 in an
Conduction II: Spreading Resistance Addendum 3

ambient temperature TA2. The new nomenclature is also shown in Figure 12.9.
The problem solution follows Section 12.6 quite closely, thus Eqs. (12.35) and (12.41) - (12.45)
apply here, but with the boundary conditions rewritten as
∂T
k − h1T = 0 at z = 0 Newtonian
∂z
∂T
(12.65)
k + h2T = 0 at z = t Newtonian
∂z
for the solution ∞ ∞
 lπ x   mπ y 
T ( r ) = ∑ ∑ ε l ε m Ψ lm ( z ) cos   cos  
l =0 m=0  a   b 
The Newtonian boundary conditions then apply to the z - dependent factor in the solution.
d Ψ lm
k
− h1Ψ lm = 0 at z = 0 Newtonian
dz
(12.66)
d Ψ lm
k + h2 Ψ lm = 0 at z = t Newtoniaan
dz
The solution will be completed for ambient temperatures of zero on both sides of the plate and will
be reintroduced via a method described in Section 12.12.
The Green’s function method is once again used to find a solution for Ψ lm .
z ′=t


1
Ψ lm = Glm ( z z ′ ) φlm ( z ′ ) dz ′ (12.67 )
k
z ′=0

Glm is the solution to d 2 Glm ( z z ′ ) 2


− γ lm Glm ( z z ′ ) = 0, z ′ ≠ z (12.68)
dz ′2

with boundary conditions


dGlm ( z z ′ )
k − h1Glm ( z z ′ ) = 0 at z ′ = 0 Newtonian
dz ′
dGlm ( z z ′ )
(12.69 )
k + h2 Glm ( z z ′ ) = 0 at z ′ = t Newtonian
dz ′
and also satisfying the remaining Green’s function properties given by Eqs. (ix.6-1) - (ix.6-8).
The general solutions for γ lm = 0 and γ lm ≠ 0 are

Glm ( z ′ z ) = Az ′ + B, z ′ < z
γ lm = 0 (12.70 )
Glm ( z z ′ ) = Cz ′ + D, z ′ > z

G ( z ′ z ) = E sinh ( γ lm z ′ ) + F cosh ( γ lm z ′ ) , z′ < z


(12.71)
G ( z z ′ ) = G sinh ( γ lm z ′ ) + H cosh ( γ lm z ′ ) , z′ > z

The Green’s function properties and Cramer’s rule are used to find the coefficients A - H to arrive at

 z  k 
 − 1  − 
t  h2 t   k
0 ≤ z ′ < z : Glm ( z ′ z ) = −   z ′ +  , γ lm = 0 (12.72 a )
  k   k   h1 
1 +   +  
  h1t   h2 t  
4 Thermal Computations for Electronics

 z ′  k 
 − 1  − 
t  h2 t   k
z < z ′ ≤ t : Glm ( z z ′ ) = −   z + , γ lm = 0 (12.72 b )
 k k   h1 
1 + + 
 h1t h2 t 

0 ≤ z′ < z :
 k γ lm 
(V sinh γ lm z − U cosh γ lm z )  sinh γ lm z ′ + cosh γ lm z ′ 
 h1 
Glm ( z ′ z ) =
 kγ  
 sinh γ lm + lm cosh γ lm z  (U sinh γ lm z − V cosh γ lm z ) 
  h1  
γ lm  
 −  cosh γ lm z + k γ lm sinh γ lm z  (U cosh γ lm z − V sinh γ lm z ) 
 
  h1  

and using reciprocity of Glm to obtain z < z ′ ≤ t , γ lm ≠ 0 (12.73)


 k γ lm 
(V sinh γ lm z ′ − U cosh γ lm z ′ )  sinh γ lm z + cosh γ lm z 
 h1 
Glm ( z z ′ ) =
 k γ lm  
 sinh γ lm + cosh γ lm z ′  (U sinh γ lm z ′ − V cosh γ lm z ′ ) 
 h 
γ lm  
1

 −  cosh γ lm z ′ + k γ lm sinh γ lm z ′  (U cosh γ lm z ′ − V sinh γ lm z ′ ) 
 
  h1  

k γ lm cosh γ lm t + h2 sinh γ lm t ,
where U = k γ lm sinh γ lm t + h2 cosh γ lm t
V=

Recalling that the surface source Fourier coefficients φlm ∝ δ ( z ′ ) use only the 0 < z ′ < z part
of Eqs. (12.72) and (12.73), the z-factors Ψ lm are found to be

 k   z  
  −  − 1 
4q  h2 t   t 
Ψ 00 = ∆x∆y (12.74 )
kabh1   k   k  
1 +   +  
  h1t   h2 t  

α l = lπ a :
 4q   ∆x∆y   kα l 
    sin (α l x2′ ) − sin (α l x1′ )  [V sinh α l z − U cosh α l z ]
 kπ lb   ∆x   h1 
Ψ lo = (12.75)
 kα l  
 sinh α l z + cosh α l z  (U sinh α l z − V cosh α l z ) − 
 h 
α l  
1

  
 cosh α l z + kα l sinh α l z  (U cosh α l z − V sinh α l z ) 
 h1  
Conduction II: Spreading Resistance Addendum 5

β m = mπ b :
 4q   ∆x∆y   k β m 
    sin ( β m y2′ ) − sin ( β m y1′ )  [V sinh β m z − U cosh β m z ]
 kπ ma   ∆y   h1 
Ψ om = (12.76 )
 k βm  
 sinh β m z + cosh β m z  (U sinh β m z − V cosh β m z ) − 
  h1  
βm  
 
 cosh β m z + k β m sinh β m z  (U cosh β m z − V sinh β m z ) 
 h1  

γ lm2 =( lπ a ) + ( mπ b ) , l ≠ 0 and m ≠ 0 :
2 2

 4q   ∆x∆y   k γ lm  
 2     sin (α l x2′ ) − sin (α l x1′ )  
 kπ lm   ∆x∆y  h1  
× sin β y ′ − sin β y ′  V sinh γ z − U cosh γ z 
  ( m 2) ( m 1 ) [ lm lm ]
Ψ lm = (12.77 )
 k γ lm  
 sinh γ lm z + cosh γ lm z  (U sinh γ lm z − V cosh γ lm z ) 
 h1 
γ lm  

 −  cosh γ lm z + k γ lm sinh γ lm z  (U cosh γ lm z − V sinh γ lm z ) 
 
  h1  

Substituting the various Ψ lm into the Fourier expansion of T (x, y, z), Eq. (12.57), and using the
definitions U, V in Eq. (12.73) with l = 0, m = 0 where appropriate, gives us

 k 
t + − z 
Q  h 
T ( x, y , z ) = 2

kab  h1t   h1  
  +   + 1
 k   h2  
 kα l   α l ∆x 
∞   sin   (V sinh α l z − U cosh α l z ) cos α l xc cos α l x
 4Q   1   h1   2 
+ 
 k π b ∆x 
∑  
l
l =1  
 kα l  
 sinh α l z + cosh α l z  (U sinh α l z − V cosh α l z ) 
 h1 
α l  

 −  cosh α l z + kα l sinh α l z  (U cosh α l z − V sinh α l z ) 
 
  h1  

 k β m   β m ∆y 
∞   sin   (V sinh β m z − U cosh β m z ) cos β m yc cos β m y
 4Q     h1   2 
1
+  ∑ 
 kπ a∆y  m =1  m   k βm  
 sinh β m z + cosh β m z  (U sinh β m z − V cosh β m z ) 
 h1 
β m  

 k β 
 −  cosh β m z + m sinh β m z  (U cosh β m z − V sinh β m z ) 
  h1  

(12.78 cont.)

α l lπ a ,=
for= β m mπ γ lm
b,= α l2 + β m2
6 Thermal Computations for Electronics

 k γ lm   α l ∆x   β m ∆y  
  sin   sin   cos α l xc cos β m yc cos α l x cos β m y 
 h1   2   2  
∞ ∞  (V sinh γ z − U cosh γ z ) 
 16Q   1 
+ 2
 k π ∆x ∆y
 ∑∑
=l 1 =
 
 lm    k γ
lm


lm



m 1
 sinh γ lm z +
lm
cosh γ lm z  (U sinh γ lm z − V cosh γ lm z ) 
  h1  
γ lm  
 γ 
 −  cosh γ lm z + k lm sinh γ lm z  (U cosh γ lm z − V sinh γ lm z ) 
  h1  

α l lπ a ,=
for= β m mπ γ lm
b,= α l2 + β m2

In order to get a formula for the thermal resistance, we set the temperature to be computed at

=x a=
2 , y b=
2 , z 0=
; xc a 2 , yc = b 2

and set the dimensionless variables as before to be


 h a  t  h t  h a  t  h t
α = ∆x a , β = ∆y a , τ = t a , ρ = a b , Bi1τ =  1    = 1 , Bi2τ =  2    = 2
 k a
  k  k  a  k
We can also decrease the computation time by noting

 lπ 
l → 2l ,then cos 2   = cos 2 ( lπ ) →1, l = 1, 2, 3 ...
 2 
 mπ 
m → 2m, then cos 2   = cos ( mπ ) →1, m = 1, 2, 3 ...
2

 2 
The dimensionless resistance is defined from

ψ = k T ( xc , yc , 0 ) Q  ∆x∆y = kR ∆x∆y


ψ = ψ U + ψ Sp , R = RU + RSp
The final two-sided cooling result is

 k   1 
t +  τ + 
∆x∆y  h2  Bi2 
ψU = = αβ ρ  (12.79 )
ab  h1t h1   Bi1 
1 + +  1 + Bi1τ + 
 k h2   Bi2 

  2lπ τ  
∞ sin ( lπ α )  tanh ( 2lπ τ ) +  
ρ β 1  Bi2τ  
ψ Sp =
π2 α ∑ l 2 

Bi1τ   Bi1τ 2lπ τ  
 tanh ( 2lπ τ ) 
l =1
 1 + + +
 Bi2τ   2lπ τ Bi2τ  
(12.80a)
  2mπ τ ρ  
∞ sin ( mπβρ )  tanh ( 2mπ τ ρ ) +  
1 α 1  Bi2τ  
+ 2 ∑
π ρ β m =1 m 2  Bi1τ   Bi1τ

2mπ τ ρ  
 1 + + +  tanh ( 2mπ τ ρ ) 
 Bi2τ   2mπ τ ρ Bi2τ  
Conduction II: Spreading Resistance Addendum 7

 sin ( lπ α ) sin ( mπβρ )    

4
∞ ∞

 2π lm l + m ρ  
2 2 2 (
  tanh 2π l 2 + m 2 ρ 2 τ ) +  2π l 2 + m2 ρ 2τ
Bi2τ


+ 2 ∑∑
π α β l =1 m =1  Bi τ   


 1 + 1  + 
Bi1τ
 Bi2τ   2π l 2 + m 2 ρ 2 τ
+
2π l 2 + m 2 ρ 2 τ
Bi2τ
(
 tanh 2π l 2 + m 2 ρ 2 τ


)

(12.80b)
Equations (12.79) and (12.80) have too many variables to allow construction of design curves in a
manageable number and although these formulae may appear a little complicated, they are not very
difficult to put into a Mathcad™ worksheet.
By this point, the reader should understand that when we use the term “uniform source,” we
mean a source that is uniformly distributed over the entire plate. In our spreading resistance problem,
this source is at z = 0 and the resistance from the uniform source plane to ambient TA2 is given by a
parallel combination RU.

1 t 1
R1 = , R2 = +
h1ab kab h2 ab
1 1 1 ab
= + = h1ab +
RU R1 R2 t 1 
 + 
 k h2 

t 1   k 
 +  t + 
1  k h2   h2 
RU = = = (12.81)
   t 1    h1t h1 
h + 1  ab  +  h1 + 1 kab 1 + + 
1
  k h2    k h2 
ab  t 1
+ 
 
 
 k h2  
 

Recasting Eq. (2.81) as a dimensionless resistance,

∆x∆y ( t + ( k h2 ) )
ψ U = kRU ∆x∆y = (12.82 )
ab 1 + ( h1t k ) + ( h1 h2 ) 

which we see is identical to Eq. (12.79).

12.12 EXPLORING THE DIFFERENCE BETWEEN ONE-SIDED AND TWO-SIDED


NEWTONIAN COOLING

The analysis of many physical problems often proceeds from the solution of the simplest ver-
sion to successively more complex models. This is the case of the single-sided Newtonian cooling
model followed by our solution of the two-sided cooling model. Equation (12.80) is rewritten as
Eq. (12.83) where factors are shown that modify the single-cooled surface to be the doubly cooled
surface. It is not surprising that the correction decreases the spreading resistance.
8 Thermal Computations for Electronics

Perhaps you will find it curious how the mathematical results change in an “organized” fashion
as a problem progresses from one level of complexity to that of another.

  Bi2τ  
  sin ( lπ α ) 1 + tanh θl   
  2 lπτ × 
 l2  Bi2τ  
∞    2lπτ + tanh θl   
ρ β α   One-sided to two-sided
ψ Sp 2 =
π 2 ∑l =1 


 1 

 cooling correction factors
 1 −  
   h2 + 1 +  Bi2τ + 2lπτ  tanh θl  
   h1   2lπτ Bi1τ 

  
 

  Bi2τ  
  1 + 2mπτρ tanh (θ m ρ )   
 ( sin m π β ρ ) × 
 m2  Bi2τ  
∞   2mπτρ + tanh (θ m ρ )   
α β    
+ ∑ 
ρ π 2 m =1  
1

 (12.83)
 1 − 
   h2 + 1 +  Bi2τ + 2mπτρ  tanh θ ρ  
  h    ( m ) 
  1   2mπτρ Bi1τ  

   
  Biτ  
 1 + tanh θlm   
  sin ( lπ α ) sin ( mπ β ρ )  2πτ l 2 + m 2 ρ 2 
 
 × 
  2π lm l 2 + m 2 ρ 2  Biτ  
  + tanh θlm   
∞ ∞   2πτ l + m ρ
2 2 2 
  
∑∑
4
+ 2  
m 1 
π α β=l 1 = 
 1 − 1 
  h   Biτ 2πτ l + m ρ
2 2 2  
   2 + 1 +  +  tanh θlm  
   h1   2πτ l 2 + m 2 ρ 2 Bi1τ  
  
   

12.13 METHOD OF INCLUDING THE EFFECT OF TWO DIFFERENT AMBIENTS


TO THE TWO-SIDED NEWTONIAN COOLING SPREADING THEORY

All of the previous spreading theory derivations have assumed a zero temperature ambient,
which works well when calculating a thermal resistance from source to ambient. With regard to the
two-sided Newtonian cooling problem, if the two ambients are equal, then the resistance from source
to ambient poses no additional considerations. Such is not the case when the two ambients are of
unequal value, but you shall now learn what to do in this situation.
The geometry and ambient temperature definitions are illustrated in Figure 12.9 where the heat
transfer coefficients h1 and h2 reference the temperatures TA1 and TA2, respectively. Suppose that we
set TA2 = 0 and TA1 = ∆TA1− A 2 + TA 2 = ∆TA1− A 2 and TA1 > TA2 (an example would be a heat
∆TA1− A 2 + 0 =
Conduction II: Spreading Resistance Addendum 9

sink where TA1 is an enclosure internal air temperture and TA2 is the room ambient temperture). Since
TA1 > TA2 will result in a greater heat sink temperature than if the two ambients were equal, we must
add the uniform source heat ∆Q to the actual source to correctly compensate.

∆Q = h1ab (TA1 − TA 2 ) = h1abTA1


We need only add the extra heat to the uniform source portion of the spreading theory because all of
the Fourier coefficients are zero for xS1 = 0, xS2 = a, yS1 = 0, yS2 = b, i.e., the issue of TA1 not equal to
TA2 does not contribute to ψ Sp !
The correction to our spreading theory begins by using the uniform source portion of Eq. (12.78).
 k 
t + − z 
( Q + ∆Q )  h2 
TU′ =
kab   h1t   h1 
1 +   +  
  k   h2 
 k 
t + − z 
( Q + h1 abTA1 )  h2  Corrected unif. (12.84 )
TU′ =
kab   h1t   h1
  temp. for TA1 > TA 2 = 0
1 +   + 

  k   h2

The modified uniform source portion of the resistance uses Eq. (12.84) with z = 0.
 h1abTA1   k 
1 +  t + 
TU′ Q h Corrected unif. (12.85 )
RU′ = =   2 

Q kab   h1t   h1   resist. for TA1 > TA 2 = 0


1 +   +   
  k   h2  
The dimensionless resistance is
 k   h1abTA1 
 t +  1 + 
∆x∆y  h2   Q 
ψ U′ = k ∆x∆yRU′ =
ab   h1t   h1  
1 +   +   
  k   h2  
 h abTA1 
αβ ρ τ + (1 Bi2 )  1 + 1 
 Q  Corrected unif. (12.86 )
ψ U′ =
(1 + Bi1τ + ( Bi1 Bi2 ) ) resist. for TA1 > TA 2 = 0

12.14 APPLICATION EXAMPLE: HEAT SINK WITH TWO CONVECTING SIDES,


ONE FINNED AND ONE FLAT

A heat sink with one finned side and one unfinned side is illustrated in Figure 12.30. The di-
mensions are: a = 4.0 in., b = 4.0 in., tb = 0.2 in., L = 1.0 in., ∆x =0.4 in., ∆y =0.4 in., Nf = 14 fins,
with other data consisting of TA1 = 0 or 30oC (above TA2 ), TA2 = 0oC, Q = 20 W, k = 5.0 W/(in. oC),
h2 = 0.01 W/(in.2 oC), and h1 = 0.0 and 0.075 W/(in.2 oC). These values of h1 and TA1 should be
adequate to show us the effects of these two variables, although the detailed calculations are shown
only for two cases, h1 = 0.0 for TA1 = 0.0oC and h2 = 0.075 and TA1 = 30oC.
Calculations of Af , RSink Fins, he , and Biot ⋅τ =he tb k are the same for both problems.

Af = 2 ( N f − 1) Lb + ab = 2 (14 − 1)(1.0 in.)( 4.0 in.) + ( 4.0 in.) = 120.00 in..2


2

RSink Fins = 1 h2 Af = 1 ( 0.01)(120.0 ) = 0.833 o C W


10 Thermal Computations for Electronics

he = 1 RSink Fins ab = 1 ( 0.833)( 4.0 in.) = 0.075 W in.2 ⋅ o C


2
( )

( Biot ⋅τ )2 = he tb k = ( 0.075 )( 0.20 ) 5.0 = 0.003, τ = 0.05


α = β = ∆x a = 0.4 in. 4.0 in. = 0.1
where the last two preceding lines are required for the calculation of the dimensionless spreading
resistance.
In the case of h1 = 0.0, we can use Figure 12.13 to obtain ψ Sp = 0.803 (the three decimal places
were obtained using a Mathcad™ calculation, not the figure). The spreading resistance is then

RSp = ψ Sp (k )
∆x∆y = 0.803 5.0
 ( 0.4 )
2  = 0.402 o C W

Now we apply Eq. (12.85) to obtain the “uniform source” contribution.

 h1abTA1   k 
1 +  t + 
RU′ = 
Q   h2  1  k  1  5.0  o
= t +  = 2 
0.2 +  = 0.836 C W
kab   h1t   h1   kab  he  ( 5.0 )( 4.0 )  0 .075 
1 +   +   
k
    2  h

The total resistance from source to ambient TA2 is then


R = RSp + RU′ = 0.402 + 0.836 = 1.238 o C W
We shall skip to the next case where h1 = 0.075 and TA1 = 0. Mathcad™ was used to calculate
ψ Sp = 0.794 . This is rather surprisingly close in value to what we obtained for the first case. Then
we obtain
RSp = ψ Sp (k )
∆x∆y = 0.794 5.0
 ( 0.4 )
2  = 0.397 o C W

 k 
t + 
(1 + h1abTA1 )  h2 
RU′ =
kab   h1t   h1 
1 +   +  
  k   h2 
 5.0 
1 + ( 0.075 )( 4.0 )2 ( 0.0 )   0.2 + 
   0.075 
RU′ = = 0.417 o C W
( 5.0 )( 4.0 )
2
 ( 0.075 )( 0.2 )  0.075  
1 + + 
 5.0  0.075  
tb

TA1 s
h1 TA2
h2
b ∆y Q
∆x
tf

a L

Figure 12.36. Application Example 12.13 : Heat sink


with two convecting sides: one finned and one flat.
Conduction II: Spreading Resistance Addendum 11

The final variation is to introduce the nonzero, source-side ambient TA1 = 30oC, for which we
need only modify RU′ .

R = RSp + RU′ = 0.397 + 0.417 = 0.814 o C W

 k 
t + 
(1 + h1abTA1 )  h2 
RU′ =
kab   h1t   h1 
1 +   +  
  k   h2 
 5.0 
1 + ( 0.075 )( 4.0 )2 ( 30.0 )   0.2 + 
   0.075 
RU′ = =15.440 o C W
( 5.0 )( 4.0 )
2
 ( 0.075 )( 0.2 )  0.075  
1 + + 
 5.0  0.075  

Thus the final result is


R = RSp + RU′ = 0.397 + 15.440 = 15.837 o C W
In summary, we see that for this problem the addition of convection to the source-side has very
little effect on ψ Sp , but a large effect on RU , so the overall result is significant. However, when we
added a large, source-side ambient temperature rise, the total resistance from source-to-ambient TA2
is increased by a large amount. This is because of the heat sink heating by TA1.

12.15 SQUARE SOURCE ON SQUARE, FINITE-MEDIA WITH ONE CONVECTING


SURFACE - TIME DEPENDENT: THEORY (RHEE AND BHATT, 2007)

Rhee and Bhatt (2007) published an analysis of the time-dependent solution of a single, centered
source on an adiabatic plane with an opposing plane where the boundary condition is Newtonian
cooling. Clearly this extends the steady-state analysis by Ellison (2003) as outlined in Section 12.6.
The geometry for the Rhee and Bhatt solution is illustrated in Figure 12.37. The appropriate partial
differential equation is

∂ 2T ∂ 2T ∂ 2T 1 ∂T α ≡ Thermal diffusivity (12.87 )


+ + = ,
∂x 2 ∂y 2 ∂z 2 α ∂t α =k ρc

Figure 12.37. Geometry for Rhee and Bhatt analysis of time-dependent


source on Newtonian cooled (one-side) substrate.
12 Thermal Computations for Electronics

With the intention of using symmetry and noting a heat source flux density q0[W/(in.2 oC)], the
boundary conditions are
z = 0 : ∂T ∂z = 0; z = c : k ∂T ∂z = − hT
x = 0, a 2 : ∂T ∂x = 0; y = 0, a 2 : ∂T ∂y = 0
(12.88)
z = 0, 0 < x < ∆x 2 , 0 < y < ∆x 2 : − k ∂T ∂z = q0
z = 0, ∆x 2 < x < a 2 , ∆x 2 < y < a 2 : ∂T ∂z = 0
t = 0 : T ( x, y , z , t ) = 0
Rhee and Bhatt calculate the temperature using Eq. (12.29), the Green’s function soultion
equation for F ( r ′ ) = 0 and QV = q0δ ( z ′ ) . They selected their three-dimensional Green’s function
GXYZ ( r , t r ′, t ′ ) as a product of one-dimensional Green’s functions from Beck et al. (1992).
t ′ ∆x 2 ∆y 2

∫∫∫
αq
T ( x, y , z , t ) = 0 GXYZ ( x, y, z , t x′, y ′, z ′ = 0, t ′ ) dy ′dx′dt ′ (12.89 )
k
τ = 0 x′ = 0 y ′= 0

GXYZ ( x, y, z , t x′, y ′, 0, t ′ ) = GX ( x, t x′, t ′ ) GY ( y, t y ′, t ′ ) GZ ( z , t 0, t ′ ) (12.90 )


t  ∆x 2   ∆y 2 

∫ ∫ ∫
αq    
T ( x, y , z , t ) = 0 GZ ( z , t 0, t ′ )  GX ( x, t x′, t ′ ) dx′  GY ( y, t y ′, t ′ ) dy ′ dt ′ (12.91)
k
τ =0  x′=0   y′=0 

“Small” time differences, α ( t − t ′ ) a 2 < 0.022 : (12.92 )


∆x 2   ∆x   ∆x  


1 ∞   na + x −   na + x + 
= GX ( x, t x′, t ′ ) dx′ ∑ erf  2  − erf  2 
2 n = −∞   4α ( t − t ′ )   4α ( t − t ′ )  
x′ = 0
     
∆y 2   ∆x   ∆x  


1 ∞   na + y −   na + y + 
= GY ( y, t y ′, t ′ ) dy ′ ∑ erf  2  − erf  2 
2 n = −∞    4α ( t − t ′ )  
4α ( t − t ′ ) 
y′=0      
1   −z2   −z2   − ( 2c − z )2  
G= ( z , t 0, t ′ )  exp   + exp   + exp  
 4α ( t − t ′ )   4α ( t − t ′ )   4α ( t − t ′ )  
Z
4α ( t − t ′ ) 

1  ( 2c − z ) α (t − t ′)   ( 2c − z ) α (t − t ′) 
− Bi exp  Bi + Bi 2  × erf  + Bi 
c 
 4α ( t − t ′ ) c2 

 4α ( t − t ′ )

c 

Rhee and Bhatt found that summation limits in Eq. (12.92) of n = -1 to +1 were satisfactory. In
this same equation, the Biot number is defined as Bi = hc/k. “Large” time differences:
∆x 2


∆x 2 ∞ 1 −4 m2π 2 α (t −t ′)  mπ x   mπ∆x 
GX ( x, t x′, t ′ ) dx′ =
a2
+ ∑ e cos   sin  
a π m =1 m  2a   a 
x′= 0 (12.93a )
∆y 2


∆x 2 ∞ 1 −4 m2π 2 α (t −τ )  mπ y   mπ∆x 
GY ( y, t y ′,τ ) dy ′ =
a2
+ ∑ e cos   sin  
a π m =1 m  2a   a 
y ′= 0
Conduction II: Spreading Resistance Addendum 13

GZ ( z , t 0, t ′ ) =
2 ∞ − βm2 α (t −t ′) c2 (β 2
+ Bi 2 ) z
∑e
m
cos  β m  (12.93 b )
c m =1 ( 2
β + Bi + Bi
m
2
 c )

The β m are the eigenvalues calculated from the roots of


β m tan β m = Bi (12.94 )
Rhee and Bhatt indicate that although the x and y-direction summations in Eq. (12.91) usually con-
verged by 10 terms, the z-direction summation converged much more slowly. In their actual numerical
studies, they used 2000 terms. Of course the actual number may differ, depending on the problem
geometry and physical constants.
The steady-state thermal resistance, total dimensionless resistance, and dimensionless spreading
resistance, respectively, are calculated from Eq. (12.91) in the usual fashion.
T ( 0, 0, 0, ∞ )
R= , ψ = k ∆xR
q0 ( ∆x )
2

 c 1 
ψ Sp = ψ −ψ U , ψ U = k ∆x  2 + 2 
 ka ha 
Rhee and Bhatt obtain steady-state results in excellent agreement with the graphical results in Sec-
tion (12.7). The interested reader is advised to consult the original paper for additional results and
clarification as desired.
14 Thermal Computations for Electronics

You might also like